Thanks. That will give me a good bit of reading. My minimalist comment
at the moment is that I would hope going to 64-bit would not be painful.
However, how do I check to see if I'm already at 64-bit? See comment
below about the actual chip I have in the PC.
I took a look at the chip. It's a Patriot PSD22G6672H. 64-bits and about
$45 from NewEgg. I presume this would be compatible with the one I have
in the PC to give me 4G, if I really have a 32-bit PC.
I like to look at the feedback, to see how many sticks arrived DOA
or had issues. I've actually used the DOA info, to change my brand
preferences.
http://www.newegg.com/Product/Product.aspx?Item=N82E16820220296
*******
If I look at the "System" control panel on my system, it says
"Microsoft Windows XP"
and there is no mention of 64 bits. So mine is 32 bits.
If you had a 64 bit version, there would likely be mention of it there.
This is what WinXP x64 bit version looks like, for its system
control panel. The "x64" tells you it is the 64 bit version
of the OS. So "no number" means 32 bit, while seeing 64 there,
means it is the 64 bit version (can handle your 8GB case).
http://www.xbitlabs.com/images/cpu/x86-64-rc1/win64-1.png
If you read the reviews for WinXP x64, drivers were one of
the issues with it. Perhaps you'd be upgrading to some
more recent 64 bit OS instead ?
http://www.newegg.com/Product/Product.aspx?Item=N82E16832116179
*******
What I'd do in your position is:
1) Buy a single 2GB stick.
2) When it arrives, you can either install it in the same
colored slot as the original (dual channel 2x2GB installed),
or run the new stick by itself while you start your testing.
3) The first thing you boot, with the new memory present, is memtest86+.
If you have the new stick by itself, there will be no ambiguity
as to which memory is causing a problem. I run this for a minimum
of two test passes. Test 5 may be the test that shows errors, if
they're going to happen. If you see just one error, in a bunch of
testing, sometimes bumping up the VDimm setting one notch, is
enough to fix it.
http://www.memtest.org/
4) Once you've completed two passes of that, your choices are then,
either to immediately boot your WinXP (32 bit) OS and try it out.
Or, boot a Linux LiveCD, as memory corruption while that is running,
won't be affecting your WinXP install. The more testing you do,
before booting WinXP, the less likely it is to get corrupted.
As long as you have a backup of the C: partition, you can likely
repair any damage in any case, whatever happens.
Memtest86+ tests all the memory, except about the first 1MB or so.
Some memory is reserved for BIOS functions, and the memory tester
can't safely test that area. Sections marked as reserved, are not
tested.
Once you've tested a single new stick on its own, that doesn't
prove that the two sticks will work together, just as well. Installing
the second stick, running 2x2GB, requires its own tests be run as
well. Some computers, become unstable when run in dual channel
mode, so it represents a separate test case.
The reason for starting with a single stick of RAM, is to make it
easier to reject the RAM right away, without investing a lot of
test time. You still need to test the "final memory configuration",
as that is the one you'll be using, and that is the one that must
be error free.
The thing is, the versions of memtest86+ plus I've used, don't say
"stick number 2 is bad". The program just gives an error address,
and it's up to your own resources, to map that to a memory stick.
And that is the main reason, for a quick test of a new stick, I
recommend testing it by itself, since then you don't have to
interpret the faulting address.
Paul